The AIoT Chat is a conversational AI assistant built directly into Ubidots. Instead of navigating dashboards, digging through device lists, and checking variables one by one, you type a question in plain language and the assistant fetches real data from your platform to answer it.
Think of it as a 24/7 data analyst for your IoT fleet: it reads what's already in your Ubidots account, analyzes it, and tells you what's happening and what you should look at next.
Requirements
Trials and Professional, Industrial, and Enterprise plans.
The AI Center is currently available to account owners and admin users. Access for end users is coming soon. |
1. How to access the AIoT Chat
From any page in your Ubidots account, click on AI Center ⟶ AIoT Chat in the navigation menu. You will land on the AIoT Chat welcome screen.
Tip: The welcome screen displays a rotating set of example queries tailored to your fleet. Click any of them to run a query instantly without typing. |
2. Using the AIoT Chat
Start a new conversation
Click + New Chat to open a fresh session. Previous conversations are saved in the sidebar so you can revisit them at any time.
Choose an agent (optional)
Next to the send button, there is a menu that lets you choose which agent handles your message. See section 4 for how to create and manage agents.
Ask a question in plain language
Type your question in the input field at the bottom. You can ask about fleet status, specific devices, variable trends, anomalies, or active alerts.
Provide context if prompted
For broad queries, the assistant may ask you to narrow the scope — for example, a specific device, time range, or variable. The more context you include upfront, the faster and more precise the response.
Review the response and follow up
The assistant returns a structured summary based on your real platform data. You can continue the conversation in the same thread to drill down into specific devices or timeframes mentioned in the response.
Note: Each conversation is independent. Context from one chat session is not carried over to a new one. |
3. What the AIoT Chat can and cannot do
The AIoT Chat is a read-only analyst. It queries and interprets the data already in your Ubidots account — it does not take actions or modify your configuration.
🟢Can do | 🔴Cannot do |
|
|
4. Agents
Agents let you customize how the AIoT Chat behaves. Instead of using a one-size-fits-all assistant, you can create specialized agents — each with its own AI model, instructions, and area of expertise — and choose which one to use in each conversation.
For example, you could have a "Predictive Maintenance Copilot" focused on anomaly detection and equipment health, and a separate "Energy Analyst" tuned to consumption patterns and efficiency reporting.
4.1. The Agents section
To manage your agents, click Agents in the AI Center navigation menu. There you'll find a searchable grid of all your agents. Each card shows the agent's name, description, and AI model.
You can search for agents by name or description using the search bar at the top.
4.2. Create an agent
Click New Agent (the + button at the upper right corner of the screen) to open the creation form. The form has two tabs:
Settings tab
Agent name (required): A human-readable name, e.g. "Solar Copilot".
Agent label: An auto-generated slug based on the name (e.g.
solar-copilot).Description: A short summary of what the agent does (up to 8,000 characters).
Model: Choose the AI model that will power the agent:
Claude Sonnet 4.6: Best balance of performance and speed for most use cases.
Claude Haiku 4.5 (Fast): Optimized for speed and lower credit consumption.
Temperature: Controls how creative or focused the agent's responses are. Range: 0–1. Lower values (closer to 0) make responses more deterministic; higher values make them more varied. Default: 0.2.
Max Tokens: Sets the maximum length of the agent's responses. Default: 4,096.
System Prompt tab
This is where you define the agent's behavior, persona, and context. Whatever you write here is sent as a system message before every conversation with this agent. Use it to describe the agent's role, areas of expertise, tone, and any constraints on how it should respond (up to 8,000 characters).
Once you're done, click Create agent.
4.3. Edit an agent
Click the the agent's card or the Explore agent button to open the same form, pre-filled with the agent's current configuration. Make your changes and click Save changes.
Note that the agent label cannot be changed after creation (even if you change the agent's name). |
4.4. Delete an agent
Click the Delete button (the trash can icon) on an agent card to remove it. A confirmation dialog will appear before anything is deleted.
A few things to keep in mind:
Deleting an agent is a soft delete — it disappears from your list immediately, but any existing conversations that used it will continue to work — the platform handles the fallback automatically.
If you delete your only agent, your account will have no agents. You can always create a new one.
4.5. Using an agent in a conversation
When typing a message in the chat, click the menu button next to the send button to select which agent should handle the conversation.
5. Usage and billing
Monthly included AI credits | 1,000 AI credits / month |
Cost per overage block | $10 USD per thousand AI credits |
Each message you send to the AIoT Chat consumes a small number of credits. The exact amount depends on three things:
How much data the assistant needs to look up: A question about a single device uses fewer credits than one that scans your whole fleet.
The length of your question and the assistant's response: Short questions and concise answers cost less.
How long your current conversation has been running: AIoT Chat re-reads the earlier messages each turn so it stays in context — that means the 10th message in a long conversation costs more than the 1st.
The monthly allowance is a soft limit. You will receive a notification as you approach it. Overage charges apply only when usage exceeds the included 1,000 AI credits.
What to expect per interaction
Type of interaction | Example | Typical credits |
Quick check | "Is pump-12 online?" | 1-2 credits |
Standard analysis | "Show me alerts from the last 24 hours" | 3-5 credits |
Multi-device or multi-variable query | "Compare temperature trends across all sites this week" | 6-10 credits |
Deep follow-up in a long conversation | The 10th+ message in the same chat | 10-15+ credits |
Heads up: Credits are shared across all conversations in your account during the billing period. Broad queries that fetch large datasets will consume more credits than targeted, specific questions. |